Re: [PATCH] Cygwin: Provide more COM devices

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

On Nov  3 20:13, Achim Gratz wrote:
> Corinna Vinschen writes:
> […]
> > ttyS%(0-255) takes another 23K btw.  Even that should be ok, if
> > the need arises.  Alternatively we could shortcut shilka as for
> > /dev/sd*, but that involved much bigger numbers.
> 
> I've searched for some documentation (anywhere the glob syntax %(1-128)
> would turn up, btw?) and I think Cygwin is misusing shilka a bit here.
> It's a keyword scanner, so the arithmetically coded parts of the device
> shouldn't be targeted at all.  Instead, only the device path prefix
> should be searched via the shilka lexer and the rest of the conversion
> done in code.

Sure, fine with me.

> For the disks we might keep the globbing that gets us the
> device major part.  That of course means we construct more devices
> on-the-fly (or even all of them) and have a much smaller table of static
> device entries

I don't think so.  We already construct the numbered devices on the
fly, see fhandler_dev::readdir() and the `exists' test in there,
which translates into on of the exists_* tests in devices.in.

> (which get searched linearly, so in the end that should
> be a net speed improvement).

That in turn would be great.


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer

Attachment: signature.asc
Description: PGP signature